EURADOS Annual Meeting 2018 - Working Group 6 Meeting Deformation of mesh-type ICRP reference computational phantoms in different statures and postures for individualized dose calculations Haegin Han Department of Nuclear Engineering, Hanyang University
New Mesh-type ICRP Reference Phantoms (MRCP) Conversion to mesh format ICRP adult voxel-type reference phantoms ICRP adult mesh-type reference phantoms One of substantial advantages is high deformability, which encourages us to deform the reference phantoms in different statures and postures. 2
Objective of the Present Study Stature deformation The stature of mesh-type ICRP reference computational phantoms (MRCPs) were deformed to represent the 10th and 90th percentiles of Caucasian population. Posture deformation A systematic procedure for realistic posture change of mesh-type phantoms was developed based on as-rigid-aspossible (ARAP) shape deformation. The developed method was used to deform the ICRP adult male mesh-type reference computational phantom (MRCP) following several postures of a real person via a motion capture device. 3
1. Phantom Deformation in Different Statures 4
10 th and 90 th percentile standing height and weight 10 th and 90 th percentile values for standing height and body weight for nine countries (i.e., Sweden, Netherlands, Germany, Belgium, Australia, USA, France, UK, and Italy) were extracted from the anthropometry program PeopleSize. 5
Determination of lean body mass Lean Body Mass (LBM) equation (Deurenberg et al. 1991) LBM = W [ W 1. 2 W H 2 + 0. 23 A 10. 8 if male 5. 4 ] 100 where LBM is Lean body mass (kg), H is height (cm), W is weight (kg), A is age (yr). Deurenberg, Paul, Jan A. Weststrate, and Jaap C. Seidell. "Body mass index as a measure of body fatness: age-and sex-specific prediction formulas." British journal of nutrition 65.2 (1991): 105-114. LBM is defined as a mass subtracting fat mass from whole body mass, and has a high correlation with the masses of most internal organs and tissues (Bosy-Westphal et al. 2004). The differences between the LBM from the ICRP Publication 89 and the LBM calculated by LBM equation for reference body shape are 0.16% for male and 0.36% for female, respectively. 6
Determination of skin mass 1. Surface area of the body(sa) was first calculated by using the equation given in ICRP publication 89. Suface area of the body SA = 0. 0235 H 0.42246 W 0.51456 where SA is surface area (m 2 ), H is height (cm), W is weight (kg). 2. Skin mass was then calculated from reference skin mass given in ICRP publication 89 by assuming that the skin mass is proportional to SA. Skin mass target = Skin mass ICRP 89 SA target SA ICRP 89 7
Determination of Anthropometric parameters Calf circumference Upper arm circumference Waist circumference Thigh circumference Buttock circumference Sagittal abdominal diameter NHANES Continuous (1999 2014) & III (1988 1994) Head breadth Head length ANSUR II (2010 2012) 8
Adjustment of phantom height 1) Head height, torso length, and leg length are scaled in the z direction, respectively, to match the target standing height. z x y 9
Adjustment of phantom height 1) Head height, torso length, and leg length are scaled in the z direction, respectively, to match the target standing height. the scaling factor for leg length (= the target leg length the leg length of adult male MRCP ) z y x 10
Adjustment of phantom height 1) Head height, torso length, and leg length are scaled in the z direction, respectively, to match the target standing height. the scaling factor for torso length (= the target torso length the torso length of adult male MRCP ) z y x 11
Adjustment of phantom height 1) Head height, torso length, and leg length are scaled in the z direction, respectively, to match the target standing height. the scaling factor for head height (= the head height at target standing height the head height at adult male MRCP height ) z y x 12
Adjustment of phantom height 2) Torso and legs are scaled in the x and y directions by using Lean body mass (LBM). z y x 13
Adjustment of Organ/Tissue Masses 2) Torso and legs are scaled in the x and y directions by using Lean body mass (LBM). y the scaling factor of the torso and arms for LBM x (= LBM target / LBM MRCP the scaling factor for torso length ) z y x 14
Adjustment of phantom height 2) Torso and legs are scaled in the x and y directions by using Lean body mass (LBM). z y x 15
Adjustment of phantom height 2) Torso and legs are scaled in the x and y directions by using Lean body mass (LBM). Dislocation problem was addressed by applying the scaling factor linearly z y x 16
Adjustment of phantom head dimension z x y 17
Adjustment of phantom head dimension x z y 18
Adjustment of phantom weight 19
Male phantoms in different body shapes H10W10-AM MRCP-AM H90W90-AM 20
Female phantoms in different body shapes H10W10-AF MRCP-AF H90W90-AF 21
Dose Calculations with Geant4 Calculated values Organ/tissue dose coefficients for photons : brain, lungs, RBM, small intestine MC simulation conditions Geant4 version: 10.04 Physics library: G4EmLivermorePhysics Secondary range cut: 1 µm Relative errors: less than 3% Photon energy: 15 kev 10 4 MeV Irradiation geometry: AP Mesh-type male phantom in Geant4 (direct implementation) 22
Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Male (10, 90%tile) - Photon beam in AP direction 10-9 10-10 Brain 10-9 10-10 Lungs 10-11 10-11 10-12 10-12 10-13 10-13 10-14 10-14 10-15 10-15 10-16 10-17 10-16 10-17 60% difference 10-18 10-9 10-10 10-2 10-1 10 0 10 1 10 2 10 3 10 4 RBM Photon Energy (MeV) 10-18 10-9 10-10 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Small Intestine Photon Energy (MeV) 10-11 10-11 10-12 10-13 25% difference 10-12 10-13 12.7% difference 10-14 11.6x difference 10-14 10-15 10-15 10-16 10-17 10-16 10-17 MRCP-AM H10W10-AM H90W90-AM 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Photon Energy (MeV) 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Photon Energy (MeV) 23
Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Female (10, 90%tile) - Photon beam in AP direction 10-9 10-10 Brain 10-9 10-10 Lungs 10-11 10-11 10-12 10-12 10-13 10-14 10-15 10-13 10-14 10-15 31.3% difference 9.4x difference 10-16 10-16 10-17 10-17 10-18 10-9 10-10 10-2 10-1 10 0 10 1 10 2 10 3 10 4 RBM Photon Energy (MeV) 10-18 10-9 10-10 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Small Intestine Photon Energy (MeV) 10-11 10-11 10-12 10-12 10-13 10-13 10-14 10-14 10-15 10-15 10-16 10-17 10-16 10-17 MRCP-AF H10W10-AF H90W90-AF 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Photon Energy (MeV) 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Photon Energy (MeV) 24
2. Phantom Deformation in Different Postures 25
As-Rigid-As Possible (ARAP) Shape Deformation CGAL doxygen (http://doc.cgal.org/latest/surface_mesh_deformation/index.html#smodelingvideo_3) 26
Principle of ARAP Shape Deformation A B Θ = 45 Rigid deformation Θ = 0 i=1 A i R i B i 2 = 0 27
Principle of ARAP Shape Deformation A B Θ = 45 ARAP deformation Θ = 0 A i R i B i 2 = E min (> 0) i=1 as similar as A i R i (rigid deformation) 28
Overall Procedure of Posture Change STEP 1 Segmentation of a mesh-type phantom STEP 2 Posture data Assign rotation matrices to corresponding segments STEP 3 A dedicated c++ program to minimize ARAP energy n E = w i M i R F 2 i=1 E min Deformed phantom STEP 4 Modify the deformed phantom to eliminate inherent defects ICRP adult male mesh-type reference phantom 29
Step 1 Phantom Segmentation The phantom was preferentially segmented into several groups according to main joints ( ). The segmented groups were also divided into halves to reduce the computation time. For the rotation, only two adjacent groups at a joint will be deformed. 30
Step 2 Assigning Rotation Matrices A rotation matrix is assigned to each segmented group considering posture change. Unit matrix Theta: 30 degree Unit vector: (-1, 0, 0) R = cosθ + n x 2 (1 cosθ) n x n y 1 cosθ n z sinθ n x n z 1 cosθ + n y sinθ n y n x 1 cosθ + n z sinθ cosθ + n x 2 (1 cosθ) n y n z 1 cosθ n x sinθ n z n x 1 cosθ + n y sinθ n z n y 1 cosθ + n x sinθ cosθ + n z 2 (1 cosθ) z θ y n x 31
Step 3 Minimization of ARAP Energy 1. ARAP energy is set for the tetrahedrons of deforming tissues (e.g., muscle, residual soft tissue, blood vessels, and lymphatic nodes), assuming that bones are rigidly rotated and the other organs are not deformed. n E = w i M i R F 2 i=1 2. The ARAP energy is minimized to deform the mesh. 1 This quadratic optimization problem can be easily transformed to a linear system. KX=d x, KY=d y, KZ=d z 2 The deformed mesh (i.e., X, Y, Z) that minimizes the ARAP energy can be obtained by using LeastSquaresConjugateGradient method provided in Eigen 3.3.90 library. 32
Phantom Deformation with ARAP Algorithm 33
Step 4 Phantom Refinement Some overlapping problems can be found, which are then refined as follows: Overlapping regions were modified, Skin layers were re-defined, Lymph nodes were re-generated, and Organs were adjusted for mass preservation. 34
Phantoms in Various Postures 35
Motion Capture System IGS-C200 Gyro Motion Capture System (Synertial) Wireless MPU Gyroscopic sensors Router Suit Battery 36
Posture Data Acquisition 37
Sitting Posture 38
Walking Posture 39
Bending Posture 40
Kneeling Posture 41
Squatting Posture 42
Phantoms deformed in different postures 43
Dose Calculations with Geant4 Calculated values Organ/tissue dose coefficients for photons w T -weighted dose coefficients for photons w T -weighted dose = w T H T = T T w T D T MC simulation conditions Geant4 version: 10.04 Physics library: G4EmLivermorePhysics Secondary range cut: 1 µm Relative errors: less than 3% Photon energy: 15 kev 10 4 MeV Irradiation geometry: AP, RLAT Mesh-type male phantom in Geant4 (direct implementation) 44
Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Photon beam in AP direction Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) w T - weighted dose per fluence (Sv cm 2 ) Absorbed dose per fluence (Gy cm 2 ) 10-9 10-10 RBM 10-9 10-10 Lungs 10-9 Colon 10-10 Colon AP 10-11 10-11 10-11 10-12 10-12 10-12 10-13 10-13 10-13 10-14 10-14 10-14 10-15 10-15 10-15 10-16 10-16 10-16 10-17 10-17 10-17 10-18 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Photon Energy (MeV) 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Photon Energy (MeV) 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) Breasts AP 10-9 Stomach 10-9 Breasts 10-9 w T - weighted dose 10-10 10-10 10-10 10-11 10-11 10-11 10-12 10-12 10-12 10-13 10-14 10-15 10-13 10-14 10-15 10-13 10-14 10-15 Standing (MRCP) Walking Sitting Bending Kneeling Squatting 10-16 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-16 10-16 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 45
Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) Photon beam in RLAT direction Absorbed dose per fluence (Gy cm 2 ) Absorbed dose per fluence (Gy cm 2 ) w T - weighted dose per fluence (Sv cm 2 ) Absorbed dose per fluence (Gy cm 2 ) 10-9 10-10 RBM 10-9 Lungs 10-10 10-9 Colon 10-10 10-11 10-11 10-11 10-12 10-12 10-12 10-13 10-13 10-13 10-14 10-14 10-14 10-15 10-15 10-15 10-16 10-16 10-16 10-17 10-17 10-17 10-18 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-18 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-9 Stomach 10-9 Breasts 10-9 w T - weighted dose 10-10 10-10 10-10 10-11 10-11 10-11 10-12 10-12 10-12 10-13 10-14 10-15 10-13 10-14 10-15 ~707x difference 10-13 10-14 10-15 Standing (MRCP) Walking Sitting Bending Kneeling Squatting 10-16 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-16 10-16 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 10-2 10-1 10 0 10 1 10 2 10 3 10 4 Energy (MeV) 46
Summary In the present study, the stature of the ICRP adult mesh-type reference computational phantom (MRCPs) were deformed to represent 10 th and 90 th percentile of Caucasian population. The adult male MRCP in several postures (i.e., walking, sitting, bending, kneeling, squatting) were also constructed by using the motion capture device and newly developed methodology based on ARAP algorithm. Photon dose coefficients (DCs) in idealized irradiation geometries for the deformed phantoms were compared with those for the original standing MRCP; significant dosimetric effects of stature/posture were found. ~12 times lower small intestine DCs were found for adult male in the 10%tile height and weight in AP at 15 kev. ~707 times lower breast DCs were found in squatting phantom in RLAT at 15 kev. 47
Summary The deformed phantoms developed in the present study are currently used by ICRP Task Group 103 to calculate dose coefficients for industrial radiography sources for use in dose estimation of workers accidentally exposed by an industrial radiography source. The developed stature/posture deformation method is semiautomatic, involving manual refinement works, which makes it exceedingly labor-intensive to construct a large number of deformed phantoms (i.e., phantom library or posture deformed phantoms for dynamic motions). We will do a further study to improve the deformation methods to be fully automatic in the future. 48
Thank you! 49
ICRP-110 Reference Phantoms Difficult to define thin or small organs Holes in hollow organs Difficult to define micron-thick radiosensitive target regions Skin: 50 µm target layer HATM/HRTM: 8-40 µm target layer Lens of eye: averagely 400 µm target layer Difficult to deform the phantoms Only upright/standing phantom with reference stature has been considered in ICRP dose calculations. ICRP adult reference voxel phantoms (ICRP Publication 110) 50
Real Exposure Situations Tokaimura nuclear accident Computed tomography A worker inspecting a pipe by industrial radiography testing Pilots in flight 51
Overall Procedure of Stature Change Stature deformation weight height 10%tile 10%tile 90%tile 90%tile Mesh-type ICRP reference phantoms 52
UF/NCI body-shape deformation methodology 1) Head, arms, and torso (including all of the internal organs and tissues) were scaled uniformly in the x, y, z direction to match the target sitting height. 2) Legs were then scaled only in the z direction to match the leg length. z y x 3) Outer body surface was deformed to match 4 anthropometric parameters (waist, buttocks, upper arm, and thigh circumference) and target weight. z 53
Overall Procedure of Stature Change 1) Head height, torso length, and leg length are scaled in the z direction, respectively, to match the target standing height. 2) Torso and legs are scaled in the x and y directions by using Lean body mass (LBM). y x 3) Head breadth and length are scaled in the x and y direction, respectively. y 4) Outer body surface was deformed to match 4 anthropometric parameters (waist, buttocks, upper arm, thigh circumferences) and target weight, but also 2 additional parameters (waist depth and calf circumference). x 54